From 0bd3f7ad5c5adf04682d48fafb40864455589890 Mon Sep 17 00:00:00 2001
From: Joshua Adams <joshua.adams@elastic.co>
Date: Thu, 25 Sep 2025 09:58:34 +0100
Subject: [PATCH] Fixes testSnapshotShutdownProgressTracker

Reorders the test to avoid a race condition due to one snapshot
thread becoming blocked

Relates: 134620
---
 .../snapshots/SnapshotShutdownIT.java         | 33 ++++++++++---------
 1 file changed, 17 insertions(+), 16 deletions(-)

diff --git a/server/src/internalClusterTest/java/org/elasticsearch/snapshots/SnapshotShutdownIT.java b/server/src/internalClusterTest/java/org/elasticsearch/snapshots/SnapshotShutdownIT.java
index 72317a7220ec9..21ab06585924c 100644
--- a/server/src/internalClusterTest/java/org/elasticsearch/snapshots/SnapshotShutdownIT.java
+++ b/server/src/internalClusterTest/java/org/elasticsearch/snapshots/SnapshotShutdownIT.java
@@ -488,6 +488,7 @@ public void testSnapshotShutdownProgressTracker() throws Exception {
         final var otherNode = internalCluster().startDataOnlyNode();
         final var otherIndex = randomIdentifier();
         createIndexWithContent(otherIndex, indexSettings(numShards, 0).put(REQUIRE_NODE_NAME_SETTING, otherNode).build());
+        indexAllShardsToAnEqualOrGreaterMinimumSize(otherIndex, ByteSizeValue.of(2, ByteSizeUnit.KB).getBytes());
         blockDataNode(repoName, otherNode);
 
         final var nodeForRemoval = internalCluster().startDataOnlyNode(
@@ -498,6 +499,7 @@ public void testSnapshotShutdownProgressTracker() throws Exception {
         final var indexName = randomIdentifier();
         createIndexWithContent(indexName, indexSettings(numShards, 0).put(REQUIRE_NODE_NAME_SETTING, nodeForRemoval).build());
         indexAllShardsToAnEqualOrGreaterMinimumSize(indexName, ByteSizeValue.of(2, ByteSizeUnit.KB).getBytes());
+        logger.info("---> nodeForRemovalId: " + nodeForRemovalId + ", numShards: " + numShards);
 
         // Start the snapshot with blocking in place on the data node not to allow shard snapshots to finish yet.
         final var clusterService = internalCluster().getCurrentMasterNodeInstance(ClusterService.class);
@@ -507,7 +509,21 @@ public void testSnapshotShutdownProgressTracker() throws Exception {
 
         waitForBlock(otherNode, repoName);
 
-        logger.info("---> nodeForRemovalId: " + nodeForRemovalId + ", numShards: " + numShards);
+        // Block on the master when a shard snapshot request comes in, until we can verify that the Tracker saw the outgoing request.
+        final CountDownLatch snapshotStatusUpdateLatch = new CountDownLatch(1);
+        final var masterTransportService = MockTransportService.getInstance(internalCluster().getMasterName());
+        masterTransportService.addRequestHandlingBehavior(
+            SnapshotsService.UPDATE_SNAPSHOT_STATUS_ACTION_NAME,
+            (handler, request, channel, task) -> masterTransportService.getThreadPool().generic().execute(() -> {
+                safeAwait(snapshotStatusUpdateLatch);
+                try {
+                    handler.messageReceived(request, channel, task);
+                } catch (Exception e) {
+                    fail(e);
+                }
+            })
+        );
+
         mockLog.addExpectation(
             new MockLog.SeenEventExpectation(
                 "SnapshotShutdownProgressTracker start log message",
@@ -553,21 +569,6 @@ public void testSnapshotShutdownProgressTracker() throws Exception {
         mockLog.awaitAllExpectationsMatched();
         resetMockLog();
 
-        // Block on the master when a shard snapshot request comes in, until we can verify that the Tracker saw the outgoing request.
-        final CountDownLatch snapshotStatusUpdateLatch = new CountDownLatch(1);
-        final var masterTransportService = MockTransportService.getInstance(internalCluster().getMasterName());
-        masterTransportService.addRequestHandlingBehavior(
-            SnapshotsService.UPDATE_SNAPSHOT_STATUS_ACTION_NAME,
-            (handler, request, channel, task) -> masterTransportService.getThreadPool().generic().execute(() -> {
-                safeAwait(snapshotStatusUpdateLatch);
-                try {
-                    handler.messageReceived(request, channel, task);
-                } catch (Exception e) {
-                    fail(e);
-                }
-            })
-        );
-
         mockLog.addExpectation(
             new MockLog.SeenEventExpectation(
                 "SnapshotShutdownProgressTracker shard snapshot has paused log message",